Separate attention backends (#3005)
This commit is contained in:
@@ -32,7 +32,7 @@ from torch import nn
|
||||
from transformers import MixtralConfig
|
||||
|
||||
from vllm.model_executor.input_metadata import InputMetadata
|
||||
from vllm.model_executor.layers.attention import PagedAttention
|
||||
from vllm.model_executor.layers.attention import Attention
|
||||
from vllm.model_executor.layers.layernorm import RMSNorm
|
||||
from vllm.model_executor.layers.linear import (LinearMethodBase,
|
||||
ReplicatedLinear,
|
||||
@@ -214,7 +214,7 @@ class MixtralAttention(nn.Module):
|
||||
base=int(self.rope_theta),
|
||||
is_neox_style=True,
|
||||
)
|
||||
self.attn = PagedAttention(
|
||||
self.attn = Attention(
|
||||
self.num_heads,
|
||||
self.head_dim,
|
||||
self.scaling,
|
||||
|
||||
Reference in New Issue
Block a user